FutA2 Is a Ferric Binding Protein from Synechocystis PCC 6803*

Synechocystis PCC 6803 has a high demand for iron (10 times greater than Escherichia coli) to sustain photosynthesis and is unusual in possessing at least two putative iron-binding proteins of a type normally associated with ATP-binding cassette-type importers. It has been suggested that one of these, FutA2, binds ferrous iron, but herein we clearly demonstrate that this protein avidly binds Fe(III), the oxidation state preference of periplasmic iron-binding proteins. Structures of apo-FutA2 and Fe-FutA2 have been determined at 1.7 and 2.7Å, respectively. The metal ion is bound in a distorted trigonal bipyramidal arrangement with no exogenous anions as ligands. The metal-binding environment, including the second coordination sphere and charge properties, is consistent with a preference for Fe(III). Atypically, FutA2 has a Tat signal peptide, and its inability to coordinate divalent cations may be crucial to prevent metals from binding to the folded protein prior to export from the cytosol. A loop containing the His43 ligand undergoes considerable movement in apo-versus Fe-FutA2 and may control metal release to the importer. Although these data are consistent with FutA2 being the periplasmic component involved in iron uptake, deletion of another putative ferric binding protein, FutA1, has a greater effect on the accumulation of iron and is more analogous to a ΔfutA1ΔfutA2 double mutant than ΔfutA2. Here, we also discover that there is a reduced level of ferric FutA2 in the periplasm of the ΔfutA1 mutant providing an explanation for its severe iron-uptake phenotype.

Iron is required for a variety of metalloproteins that play central roles in fundamentally important biological processes, including photosynthesis and respiration. Under aerobic conditions the thermodynamically favored oxidation state of iron is Fe(III), which has limited bioavailability due to its insolubility in water at neutral pH. Organisms have therefore developed various ways to handle iron that circumvent the limitations imposed by its inorganic chemistry (1)(2)(3)(4)(5)(6)(7)(8). In mammals, transferrin plays a key role in binding and solubilizing Fe(III) for import into cells (1,3,7). Various microorganisms produce siderophores, which are small organic chelating ligands that facilitate Fe(III) uptake (2, 4, 6 -8). Bacteria also possess transferrin-like molecules, termed ferric binding proteins (Fbps) 4 (4, 5, 9 -12), which form part of ATP-binding cassette-type importers (13)(14)(15)(16). Analogous systems exist for a range of metals, and in all cases they possess a metal-binding protein either in the periplasm or attached to the plasma membrane (17)(18)(19).
Cyanobacteria such as Synechocystis PCC6803, which have major metal requirements (20,21), provide good models for the analysis of metal-ion transport by ATP-binding cassettetype importers, because they possess systems for iron, manganese, and zinc in the periplasm (16,18,19,22,23). The iron transport system, termed Fut, has been proposed to comprise the four proteins FutA1, FutA2, FutB, and FutC (16). The suggested functional scheme (16) unusually has two putative periplasmic Fbps, FutA1 and FutA2, which share 52% sequence identity and have apparently redundant roles. FutB and FutC are subunits of the plasma membrane iron transporter complex (the permease and ATPase, respectively). Consistent with this proposal is the finding that the ⌬futA1 and ⌬futA2 mutations in Synechocystis result in reduced iron uptake, with the former having a much more significant effect (37 and 84% residual uptake, respectively, when compared with the wild-type strain (12)). The ⌬futA2 mutation also affects copper supply to plastocyanin in the thylakoid, and the absence of FutA2 has been suggested to lead to iron association at sites that block copper binding (24). Proteomic studies reveal that FutA2 is one of the most abundant soluble proteins in the periplasm (25) (some FutA2 has been observed in Synechocystis membrane preparations and also in the cytoplasm (26 -28)). The N-terminal region of FutA1 possesses a putative "lipoprotein" motif sequence (29), which may allow localization to the outer surface of the plasma membrane where it could function as part of the iron import system. However, subcellular localization studies have shown that FutA1 is predominantly an intracellular protein, and the phenotypes of deletion mutants suggest a function * This work was supported in part by Newcastle University and Biotechnology and Biological Sciences Research Council Grants BB/E016529 and BB/E001688/1. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. in the protection of photosystem II under conditions of iron limitation in Synechocystis and also Synechococcus strains (21,26,30). Consistent with this role, FutA1 has been observed in both thylakoid membrane (26,27) and plasma membrane (28) preparations (some material has also been found in the outer membrane (32)). To date, FutA1 has not been identified in the soluble fraction of the periplasm (24,26,27). Fbps have been classified on the basis of sequence and structure, including known iron coordination environments (33). The so-called class III members seem not to possess a small molecule anionic ligand, a feature thought to be common to all other classes of Fbps (11,(33)(34)(35)(36) and mammalian transferrins (3,37). An exception is the Fbp from Yersinia enterocolitica, which has five protein-derived ligands and a coordinated water molecule (38). It has been suggested that the class III proteins represent the earliest forms of Fbps from which proteins with modified iron sites and different synergistic anion-binding requirements have evolved (33). FutA1 and FutA2 have been assigned to this class, and for the former this has been confirmed by the crystal structure of the protein (36).
The class III Fbp from Campylobacter jejuni (cFbpA) reportedly binds Fe(II), which is subsequently oxidized to Fe(III) (33). It has recently been suggested that both FutA1 and FutA2 favor Fe(II), with hardly any detectable Fe(III) binding and that this preference is likely to extend to all class III Fbps (36). However, our previous in vitro analysis of FutA2 (24) and that of others with FutA1 (12) imply that both proteins bind Fe(III). If FutA2 does bind divalent metal ions, then the model (24) for how this protein influences copper uptake would need to be reconsidered. Given the potentially different subcellular localization of FutA1 (intracellular) and FutA2 (periplasmic) and the emerging uncertainty surrounding iron/metal oxidation state (charge) preference of this and other class III Fbps, further studies on FutA2 were required. These have included detailed Febinding investigations and determination of the crystal structures of both the apo-and Fe-loaded protein, which reveal that FutA2 is indeed a member of the anion-independent class III Fbps. Under rigorously anaerobic conditions FutA2 binds very little Fe(II) and preferentially binds Fe(III) via four tyrosines and a histidine. The charge at and around the metal site favors the binding of a trivalent cation. Fractionation of periplasm extracts from ⌬futA1 cells, followed by quantification of iron via inductively coupled plasma-mass spectrometry, reveals an unexpected interaction between FutA1 and the accumulation of ferric FutA2.

MATERIALS AND METHODS
Isolation and Purification of FutA2-Iron-loaded FutA2 was isolated and purified as described previously (24). The iron-free (apo) form of the protein was prepared by incubating Fe-FutA2 in a 3000-to 6000-fold molar excess of sodium citrate in 25 mM Tris (24). Apo-FutA2 prepared in this way contained Ͻ10% iron, as indicated by the intensity of the UV-visible band (spectra were measured at 25°C on a 35 spectrophotometer; PerkinElmer Life Sciences) at 450 nm (⑀ ϭ 6 ϫ 10 3 M Ϫ1 cm Ϫ1 ) as compared with the band at 280 nm (the A 280 /A 450 ratio is 11.5 for pure Fe-FutA2). This was confirmed by determining iron concentrations with an M Series atomic absorption spectrometer (Thermo Electron Corp.).
Iron Binding Experiments-Iron binding by apo-FutA2 was investigated by UV-visible spectrophotometry, monitoring the increase in absorbance at 450 nm upon the addition of ferric chloride and ferrous ammonium sulfate. The iron concentrations of stock ferrous and ferric solutions were determined by atomic absorption spectrometry. The amount of Fe(II) in these solutions was assayed using the strong ferrous chelator 1,10phenanthroline (phen) making additions from a 0.1 M stock solution in ethanol (final concentration, 1 mM). The formation of the [Fe(phen) 3 ] 2ϩ complex was monitored at 510 nm, and the Fe(II) concentration was calculated as described previously (⑀ ϭ 1.1 ϫ 10 4 M Ϫ1 cm Ϫ1 ) (36,39,40). Stock solutions of Fe(III) contained Ͻ 1% Fe(II) and no additional Fe(II) formed following dilution in buffer (usually 5 mM Tris plus 200 mM NaCl at pH 8.0). Aerobic iron binding experiments were performed by titrating an iron solution (1 mM in water for Fe(II) and 2 mM in 20 mM sodium citrate for Fe(III) (40)) into the apo-protein (15-25 M) in 5 mM Tris plus 200 mM NaCl at pH 8.0. Mixtures were incubated at 25°C for 15-30 min prior to acquisition of a UV-visible spectrum. Control experiments testing the stability of Fe(II) were performed by incubating aliquots of the metal in buffer (with or without 10 M sodium citrate). After different incubation times, phen (1 mM) was added, and the concentration of the [Fe(phen) 3 ] 2ϩ complex was calculated as described above. For anaerobic Fe(II) titrations apo-FutA2 (15 M) was in 20 mM Hepes with 200 mM NaCl plus 0.2 mM sodium ascorbate at pH 7.0 and was incubated overnight in an anaerobic chamber (Belle Technology). The solution was transferred to a gastight quartz cuvette, an Fe(II) solution (2 mM in 20 mM sodium ascorbate prepared under anaerobic conditions) was titrated into the apo-protein, and the mixture was allowed to equilibrate (20 min). The cuvette was then removed from the anaerobic chamber, and a UV-visible spectrum was recorded. FutA2 was desalted on a PD10 column (GE Healthcare) under anaerobic conditions in 20 mM Hepes plus 200 mM NaCl at pH 7.0.
Crystallization and Data Collection-Purified Fe-FutA2 was concentrated to 15 mg/ml in 5 mM Tris at pH 7.5 and subjected to extensive screening to identify conditions supporting crystallization, using commercially available solutions. Diffractionquality colorless crystals, presumed to be of the apo-protein, took a few months to grow. These crystals were obtained using sitting drops at 20°C in 0.5 M ammonium sulfate, 1 M lithium sulfate, and 0.1 M tri-sodium citrate at pH 5.6 (citrate removes Fe(III) from the protein) from a screen set-up with a crystallization robot using 100 nl of protein and 100 nl of precipitant solution. Diffraction-quality crystals that remained intensely red/brown, presumed to be of the Fe-loaded protein, were grown in hanging drops using 1 l of protein and 1 l of precipitant at 20°C in 27% polyethylene glycol 4000, 0.2 M magnesium chloride, and 0.1 M Mes at pH 6.5. Again, the crystals took a few months to grow. Prior to data collection, crystals of apoand Fe-FutA2 were cryoprotected by immersion in N-paratone oil and frozen by plunging into liquid nitrogen. X-ray diffraction data for apo-FutA2 were collected on station 10.1 at the Daresbury Synchrotion Radiation Source (operating at ϭ 0.979 Å) with a Mar-CCD detector and the crystal maintained at 100 K using a gaseous nitrogen cold stream. Data for Fe-FutA2 were collected using x-rays from a Rigaku Micromax007 generator and R-Axis IV ϩϩ detector. The crystal was maintained at 93 K as above. All data were integrated with MOSFLM (41) and merged/scaled with SCALA (42) as implemented in the CCP4 suite (43). 5% of data were set aside for calculation of R free . Data collection and processing statistics are given in Table  1. Assuming a solvent content of 36% (apo-FutA2) and 50% (Fe-FutA2), the asymmetric unit of each crystal contains two molecules.
Structure Solution and Refinement-The structure of apo-FutA2 was initially solved by molecular replacement using a 1.8-Å resolution data set collected on a home source (described above) by dividing the structure of the apo form of the Fbp from Mannheimia hemolytica (also known as Pasteurella hemolytica, PDB code 1Q35 (44)) into its N-and C-terminal domains for separate rotation/translation searches. PHASER (45) was used to position each domain resulting in two complete molecules. The protein sequence was corrected, and an initial model was built in COOT (46). This structure was used to solve the synchrotron data set, and a final model was produced by repeated cycles of model building in COOT and refinement with REFMAC (47), as implemented in the CCP4 suite. Attempts to solve the Fe-FutA2 structure using the complete apo-FutA2 model did not produce a clear solution. Splitting the model into separate N-and C-terminal domains significantly improved structure solution, indicating that a sizable domain rearrangement occurs in FutA2 upon iron binding. The final Fe-FutA2 model was produced by repeated model building and refinement cycles as described for apo-FutA2. Refinement statistics for apo-and Fe-FutA2 are given in Table 1. The apo-FutA2 model comprises 4897 non-hydrogen protein atoms, 2 sulfate anions, and 799 water molecules, whereas the Fe-FutA2 model comprises 4876 non-hydrogen protein atoms, 2 iron atoms, 2 chloride anions, 2 molecules of Mes, and 22 water molecules. For analysis, MOLPROBITY (48) and LSQMAN (49), respectively, were used to generate Ramachandran plots and superimposed structures from which root mean square deviation (r.m.s.d.) values based on C ␣ atoms were determined.
Surface area calculations were performed with PISA (50), and protein structure figures were prepared with PyMOL. 5 The coordinates and structure factors for apo-FutA2 and Fe-FutA2 have been deposited with the Protein Data Bank.
Analysis of Periplasm Extracts-Periplasm contents were liberated from Synechocystis PCC 6803 via cold osmotic shock resolved by two-dimensional liquid chromatography, and then analyzed for iron by inductively coupled plasma-mass spectrometry and for proteins by SDS-PAGE as described previously (24). The amount of iron in each fraction is represented as a surface rather than as contours.

RESULTS
FutA2 Acquires Ferric and Not Ferrous Iron-The binding of iron by apo-FutA2 has been analyzed under aerobic and rigorously anaerobic conditions monitoring the ligand to metal charge transfer band at 450 nm in the visible spectrum, a feature characteristic of Fbps (3,12,24,33,34,36,52) and associated with the coordination of Fe(III) by the phenolates of Tyr residues. The ferric chloride solution (Ͻ1% Fe(II) with no additional ferrous formation following dilution in 5 mM Tris plus 200 mM NaCl at pH 8.0 for the duration of the titration) contained a 10-fold excess of sodium citrate to stabilize this oxidation state of the metal (40). Under these conditions apo-FutA2 binds 1 equivalent of Fe(III), as shown in Fig. 1 and 2 (similar results were obtained when the ferric ion was stabilized in methanol (53), data not shown). The addition of Fe(II) to apo-FutA2 in the same buffer also results in the incorporation of one equivalent of iron (Fig. 2) as reported previously (36). A control experiment in which the ferrous ammonium sulfate stock solution was diluted in buffer alone, demonstrated that the Fe(II) concentration decreases rapidly with time due to auto-oxidation (54,55). It has been shown that this process is accelerated by increasing pH (54) and also by the presence of oxygen containing Fe(III) chelators, such as citrate and phosphate (54,55). We found that low citrate concentrations (10 M) decreases the Fe(II) half-life and anticipate that apo-FutA2 will have an even  more significant effect due to its enhanced Fe(III) binding capability. We have therefore studied the binding of Fe(II) by apo-FutA2 under strictly anaerobic conditions, at pH 7.0 in Hepes buffer, and in the presence of ascorbate, to endeavor to maintain iron in the ferrous form (40). Under these conditions, Ͻ10% of the protein acquired iron, as judged from the increase in absorbance at 450 nm, after exposure to three equivalents of Fe(II) for Ͼ1 h (Fig. 2). Removal of the excess Fe(II) on a desalting column yielded a small increase in the amount of ironloaded FutA2 (ϳ10%) as determined by atomic absorption spectrometry.
Overall Structures of Apo-and Fe-FutA2-The FutA2 construct used in these studies codes for residues 32-346 plus an N-terminal Met, which was maintained in the overexpressed protein (experimental mass of 35002 Da determined by matrixassisted laser desorption ionization time-of-flight mass spectrometry, compared with a theoretical value of 34,998 Da). The crystals of apo-and Fe-FutA2 each contain two molecules in the asymmetric unit. These two molecules overlay with r.m.s.d. values of 0.27 Å and 0.32 Å for apo-and Fe-FutA2, respectively (based on 313 equivalent C ␣ atoms), demonstrating that for comparison purposes they are essentially identical. Overall, the structure of FutA2 is very similar to those of other Fbps (11,35,38) and in particular the class III members FutA1 (36) and cFbpA (33). Structure overlays give the following r.m.s.d. values: 1.09 Å (275 equivalent C ␣ atoms) for Fe-FutA2 and Fe-cFbpA, 1.12 Å (300 equivalent C ␣ atoms) for apo-FutA2 and apo-FutA1, and 1.28 Å (173 equivalent C ␣ atoms) for Fe-FutA2 and Fe-FutA1. FutA2 is a two-domain protein, adopting the well characterized Fbp fold (SCOP (56) domain number 53867) with the iron binding site in a cleft between the two domains (see Fig. 3A). Residues from both domains contribute to metal binding. The N-terminal domain comprises residues 34 -129 and 263-311, whereas the C-terminal domain consists of residues 130 -262 and 312-346. The hinge between the domains is formed from two ␤-strands, which allow for flexibility in this region (33,36). The overall buried surface area between the domains (as described above) is 1130 and 1063 Å 2 for apo-FutA2 and Fe-FutA2, respectively.  (Fig. 3B). The four coordinating O atoms are all ϳ2.0 -2.2 Å from the iron, and there is very little difference between the metal sites in the two molecules of the asymmetric unit. The angles between the atoms forming the trigonal plane (Fe(III) and O atoms of Tyr 44 , Tyr 169 , and Tyr 226 ) range from 94°to 134°, whereas the angle between the axial ligands is close to 180°. The iron site geometry is therefore best described as distorted trigonal bipyramidal. There are only subtle differences at the Fe(III) site of FutA2 compared with those of FutA1 (36) and cFbpA (33). The most significant difference is that in FutA1 the iron is 2.56 Å from the N ⑀2 atom of the His 54 ligand (36).

The Iron-binding Environment of FutA2 Is Optimal for Ferric Ions-Fe(III) is bound at the domain interface of
The positions of the Tyr ligands in Fe-FutA2 are stabilized by residues that surround the coordinating amino acids. Arg 129 forms the most important interactions, hydrogen bonding via its N 2 atom to the O of Tyr 44 , and with the O atoms of Tyr 169 and Tyr 225 through its N 1 moiety. These hydrogen bonds, except for a weak interaction between Arg 129 and Tyr 225 , are absent in the structure of the apo-protein, which may be due to the Tyr residues being protonated. There are a number of other charged side chains around the metal-binding site of FutA2, some of which help to stabilize the iron-loaded form. Arg 42 and Glu 300 form intimate hydrogen-bonding contacts that are coupled with the re-orientation of His 43 upon iron binding (see below). Asp 47 , more distant from the metal site, hydrogen bonds with Arg 42 in both Fe-and apo-FutA2, and the side chain  of Glu 300 also hydrogen bonds to the N ⑀ atom of Arg 129 . Additionally, the surface-exposed Arg 229 is linked via Asn 298 to Glu 300 , and the side chains of Asp 205 and Arg 131 interact via hydrogen bonds. These, and numerous other interactions, provide a rich second coordination sphere in FutA2, and remarkably most of this is almost identical in FutA1 (36). In fact, of the 27 residues that have atoms lying within a 10-Å radius of the metal (not including the ligands) 23 are conserved in the two proteins. In the corresponding region of cFbpA (33) 15 of these 27 residues are conserved.
Movement of His 43 May Facilitate Iron Release-Iron binding to FutA2 results in both global re-arrangement of the domains (Fig. 4A) and changes in the relative positions of individual residues to accommodate the metal (Fig. 4B). The hinge (made up of two ␤-strands) allows the domains to act as a clamp upon iron binding. Interestingly, Asp 87 , a residue not on either strand but located on the "elbow" of the hinge, is the only significant outlier in the Ramachandran plot for apo-FutA2. The strained conformation of this residue relaxes when iron binds, and a similar feature has been observed (36) in FutA1 (for Asp 103 ).
The short ␣-helix (Asp 288 -Gly 297 ) adjacent to the hinge region of FutA2 shifts significantly upon iron binding. This provides space for the loop containing residues His 43 and Tyr 44 to be positioned such that their side chains can bind the metal. When considering the iron center as being made up of two "half-sites," there is little change in the position of the C-terminal Tyr ligands (Tyr 169 , Tyr 225 , and Tyr 226 ) upon iron binding (Fig. 4B). However, there are significant side-chain re-arrangements for the coordinating residues in the N-terminal domain, which result in a shift of 1.6 Å for the O atom of Tyr 44 and 10.9 Å for the N ⑀2 atom of His 43 (Fig. 4B). In apo-FutA2 the position of His 43 is stabilized by a hydrogen bond between its N ⑀2 atom and the O ␦1 atom of Asp 48 . This interaction may contribute to the somewhat strained conformation of residues 45-47 in apo-FutA2, which results in ambiguous electron density for this region. In contrast, this area is well defined in Fe-FutA2.
Amounts of Fe-FutA2 Are Low in the ⌬futA1 Periplasm-Evidence that iron uptake is more severely compromised in a ⌬futA1 than a ⌬futA2 mutant (12) appears to conflict with a role for FutA2 as the main periplasmic Fbp. Periplasm extracts were therefore prepared from a ⌬futA1 mutant and analyzed for the abundance of iron-FutA2 via two-dimensional liquid chromatography in a similar manner to previous assays of wild-type and ⌬futA2 strains (24). The FutA2 iron pool was substantially lower in ⌬futA1 extracts (Fig. 5A). Some FutA2 is detected by SDS-PAGE (Fig. 5B), although the band intensity relative to other periplasmic proteins was noticeably less than that previously observed for wild type (24) and than that seen here following SDS-PAGE of ion exchange fractions from both genotypes (Fig. 5B). The FutA2 band is, of course, absent altogether in the ⌬futA2 mutant. However, the total amount of iron in other protein and low molecular weight complexes was ϳ3-fold higher in periplasmic extracts from ⌬futA1 and ⌬futA2 relative to wild type ( Table 2). These observations are consistent with loss of iron import at the plasma membrane leading to an accumulation of iron in the periplasm.

DISCUSSION
The presence of a Tat export signal (26) implies that FutA2 is folded in the cytosol prior to export to the periplasm (57). Once folded the protein would presumably be able to bind iron, although the extent to which FutA2 will have access to freely available iron in the cytosol is not known. Metal binding by FutA2 prior to export seems to be inconsistent with a role for the protein in iron import. We clearly demonstrate herein that FutA2 preferentially binds Fe(III) over Fe(II). Apparent Fe(II) binding is seen under aerobic conditions due to the instability of Fe(II), which rapidly oxidizes to Fe(III) (58). Under conditions where Fe(II) is stabilized, limited iron binding is observed, even after extended periods of incubation. Therefore, Fe(III), and not Fe(II), binding is monitored in aerobic experiments using Fe(II), as seen previously for FutA1 (36) and cFbpA (33). This does not indicate that these proteins actually favor Fe(II) binding. Preferential binding of Fe(III) can be rationalized electrostatically considering that the four phenolate ligands present a 4Ϫ charge at the metal site of FutA2, with the guanidinium group of Arg 129 neutralizing one of these. Total charge balance therefore occurs upon binding a 3ϩ cation at this site. Preferential binding of Fe(III) is a common feature of all Fbps, as exemplified by the observation that the protein from Neisseria meningitides, which binds a synergistic anion (class I), has an Fe(III) affinity 12 orders of magnitude larger than that for Fe(II) (58). This is consistent with the iron binding properties observed for mammalian transferrins, which display a preference of 17 orders of magnitude for Fe(III) (3,7,59). Importantly, FutA2 can fold and exist as a stable apo-form and then subsequently acquire metal. Thus, even if FutA2 is exported as a folded protein, the strong preference for a trivalent cation could help preclude metal binding, including Fe(II), to folded FutA2 in the cytoplasm, allowing it to acquire ferric ions in the periplasm.
The observations made here about the iron oxidation state preference of FutA2 must also apply to FutA1, due to its almost identical metal-binding environment, which extends to the second coordination sphere. FutA1 has not been identified in the soluble fraction of the periplasm (24,25), which argues against a role in iron transfer to a plasma membrane-based importer. The studies that have suggested that FutA1 functions intracellularly to protect photosystem II during iron deficiency (21,26,30) have not determined the molecular details of this interaction (possibly interacting with the non-heme iron of photosystem II). However, given that iron homeostasis, oxidative stress, and redox regulation are closely linked (21) this function of FutA1 may require Fe(III) binding. It has been proposed that FutA1 could have acquired a different/ additional task during evolution but has retained significant homology to the family of its original function (26). FutA2 is the most abundant soluble iron-binding protein in the periplasm of Synechocystis (24). It therefore appears likely that FutA2 is the primary Fbp acting in the iron-uptake system in this organism. This conclusion seems to contradict the finding that the ⌬futA1 mutation has a much more significant effect than ⌬futA2 on iron uptake (12). The severely impaired iron-uptake phenotype of the ⌬futA1 mutant (12) can, in part, be explained by a coincident decline in Fe-FutA2 within the periplasm of this strain (Fig. 5). Consequently the amount of iron that backs up in the periplasm is similar in the ⌬futA1 and ⌬futA2 mutants (see Table 2). This raises further questions about the role of FutA1 in the regulation of production, export, and/or iron status of FutA2.
It is interesting to note that the Synechocystis genome possess a third gene that has been annotated as an Fbp (slr0237) and has been assigned to class V (33). This protein only shares two of the five metal coordinating residues found in FutA2 (and FutA1), and therefore the way in which it binds iron (if it does at all) is likely to be different. It remains to be determined if this protein has a role in iron homeostasis.
The Fe(III) binding site of FutA2 is found at the interface between the two mainly ␣-helical, and structurally similar, subunits with Tyr and His ligands from the N-terminal domain and three coordinating Tyr residues from the C-terminal domain. The iron site does not possess any exogenous anions as ligands, which is consistent with the suggestion that this unusual feature is common to all class III Fbps (36). The distorted trigonal bipyramidal geometry is also common to the class III proteins, whereas those Fbps with anion-binding metal sites usually possess a six-coordinate, approximately octahedral, arrangement around the iron (11,34). A notable exception is FbpA from M. hemolytica, which has a five-coordinate site with three Tyr ligands and a bidentate synergistic carbonate anion (35) (the Y. enterocolitica protein has an octahedral site, including a coordinated water (38)).
The two domains of FutA2 are linked by a flexible hinge, which allows them to clamp onto the metal ion when it binds. The global changes observed upon iron binding/release are similar to those for FutA1 (36), and the FbpAs from M. hemolytica (35,38), and H. influenzae (11,51). These alterations are analogous to those seen in mammalian transferrins, although the domain shift in Fbps is not as extreme (3). The most notable alteration in the immediate vicinity of the metal site of FutA2 is for the N-terminal loop on which the His 43 and Tyr 44 ligands are located. In particular, the side chain of His 43 is found in a completely different position in apo-and Fe-FutA2, suggesting this residue plays a significant role in the iron release mechanisms (both N-terminal ligands could be involved). FutA2 loses metal when the pH is lowered (data not shown), and His 43 appears to be protonated in the apo-FutA2 structure (crystals obtained at pH 5.6). Displacement of His 43 upon contact with the FutBC complex may thus act as a trigger to aid iron release. Such a mechanism resembles that proposed for metal ion release from the copper metallochaperone Atx1 to the thylakoid copper importer PacS in the same organism (31). It should not be overlooked that Tyr ligand protonation could facilitate iron release and that iron reduction may also play a role.

CONCLUSIONS
The structures of apo-and Fe-FutA2 show that FutA2 coordinates Fe(III) in a trigonal bipyramidal arrangement without the assistance of a synergistic anion. Movement of the His 43 FIGURE 5. Negligible Fe-FutA2 in the ⌬futA1 periplasm. A, periplasm extracts of ⌬futA1 (bottom) and wild type (top) were resolved by two-dimensional anion exchange and high pressure size exclusion chromatography. Fractions were analyzed for iron by inductively coupled plasma-mass spectrometry, and the lower z-axis covers a range 5.2 times greater than the upper. The inset shows the Fe-FutA2 region for both genotypes (open symbols, ⌬futA1; closed symbols, wild type) plotted on the same scale. Similar profiles were obtained with three independent extracts, as shown previously for wildtype (24). B, fractions, post ion exchange chromatography for all three genotypes, were resolved by SDS-PAGE (left), and proteins were visualized using Sypro Ruby. The prominent boxed band (wild type) has been excised and identified as FutA2 via peptide mass fingerprinting. Aliquots of fractions indicated with a horizontal bar in the ⌬futA1 profile following the second dimension of gel filtration chromatography were resolved by SDS-PAGE (right). ligand may play an important role in iron release. Iron binding experiments demonstrate that FutA2 has a preference for Fe(III), which is consistent with the charge at and around the metal-binding site. This could ensure that folded FutA2 is exported without metal allowing it to bind ferric ions in the periplasm. It is therefore likely that FutA2 is the main Fbp involved in binding and transferring Fe(III) in the periplasm of Synechocystis. Unexpectedly, the accumulation of periplasmic Fe-FutA2 shows an intriguing dependence upon cytosolic FutA1, and the underlying mechanism requires further investigation.